linux-kernel - Re: [RFC PATCH 2/8] mm/damon/core: implement intervals auto-tuning

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-Id: <20250225181449.3008-1-sj@kernel.org>
Date: Tue, 25 Feb 2025 10:14:49 -0800
From: SeongJae Park <sj@...nel.org>
To: SeongJae Park <sj@...nel.org>
Cc: Andrew Morton <akpm@...ux-foundation.org>,
	damon@...ts.linux.dev,
	kernel-team@...a.com,
	linux-kernel@...r.kernel.org,
	linux-mm@...ck.org
Subject: Re: [RFC PATCH 2/8] mm/damon/core: implement intervals auto-tuning

On Wed, 12 Feb 2025 17:44:32 -0800 SeongJae Park <sj@...nel.org> wrote:

> Implement the DAMON sampling and aggregation intervals auto-tuning
> mechanism as designed on the cover letter of this patch series.  The
> mechanism reuses the feedback loop function for DAMOS quotas
> auto-tuning.  Unlike the DAMOS quotas auto-tuning use case, limit the
> maximum decreasing amount after the adjustment to 50% of the current
> value.  This is because the intervals have no good merits at rapidly
> reducing, and it is assumed the user will set the range of tunable
> values not very wide.
> 
> Signed-off-by: SeongJae Park <sj@...nel.org>
> ---
>  include/linux/damon.h | 16 ++++++++++
>  mm/damon/core.c       | 68 +++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 84 insertions(+)
[...]
> diff --git a/mm/damon/core.c b/mm/damon/core.c
> index 2fad800271a4..227bdb856157 100644
> --- a/mm/damon/core.c
> +++ b/mm/damon/core.c
[...]
> +static unsigned long damon_get_intervals_adaptation_bp(struct damon_ctx *c)
> +{
> +	struct damon_target *t;
> +	struct damon_region *r;
> +	unsigned long nr_regions = 0, access_samples = 0;
> +	struct damon_intervals_goal *goal = &c->attrs.intervals_goal;
> +	unsigned long max_samples, target_samples, score_bp;
> +	unsigned long adaptation_bp;
> +
> +	damon_for_each_target(t, c) {
> +		nr_regions = damon_nr_regions(t);

This should use '+=' instead of '='.

> +		damon_for_each_region(r, t)
> +			access_samples += r->nr_accesses;
> +	}
> +	max_samples = nr_regions * c->attrs.aggr_samples;
> +	target_samples = max_samples * goal->samples_bp / 10000;
> +	score_bp = access_samples * 10000 / target_samples;
> +	adaptation_bp = damon_feed_loop_next_input(100000000, score_bp) /
> +		10000;
> +	/*
> +	 * adaptaion_bp ranges from 1 to 20,000.  Avoid too rapid reduction of
> +	 * the intervals by rescaling [1,10,000] to [5000, 10,000].
> +	 */
> +	if (adaptation_bp <= 10000)
> +		adaptation_bp = 5000 + adaptation_bp / 2;
> +
> +	return adaptation_bp;
> +}
[...]
> @@ -2204,6 +2262,8 @@ static void kdamond_init_intervals_sis(struct damon_ctx *ctx)
>  	ctx->next_aggregation_sis = ctx->attrs.aggr_interval / sample_interval;
>  	ctx->next_ops_update_sis = ctx->attrs.ops_update_interval /
>  		sample_interval;
> +	ctx->next_intervals_tune_sis = ctx->next_aggregation_sis *
> +		ctx->attrs.intervals_goal.aggrs;
>  
>  	damon_for_each_scheme(scheme, ctx) {
>  		apply_interval = scheme->apply_interval_us ?
> @@ -2290,6 +2350,14 @@ static int kdamond_fn(void *data)
>  		if (ctx->passed_sample_intervals >= next_aggregation_sis) {
>  			ctx->next_aggregation_sis = next_aggregation_sis +
>  				ctx->attrs.aggr_interval / sample_interval;
> +			if (ctx->attrs.intervals_goal.aggrs &&
> +					ctx->passed_sample_intervals >=
> +					ctx->next_intervals_tune_sis) {
> +				ctx->next_intervals_tune_sis +=
> +					ctx->attrs.aggr_samples *
> +					ctx->attrs.intervals_goal.aggrs;
> +				kdamond_tune_intervals(ctx);
> +			}

kdamond_tune_intervals() may increase ctx->next_aggregation_sis inside
damon_set_attrs().  So it should be called before the above
ctx->next_ops_update_sis update.

>  
>  			kdamond_reset_aggregated(ctx);
>  			kdamond_split_regions(ctx);
> -- 
> 2.39.5
> 


Thanks,
SJ